Technology and Science14 July, 20263 min read
Adam Mosseri Says Meta May Cap AI Token Spending Per Engineer Within Two Years
This article is an automated aggregation of coverage from news outlets. Our editorials offer in-depth analysis curated by the NewsCord team.
Meta may cap AI token spending per engineer within one to two years. Cap would reflect AI costs potentially equaling or exceeding salaries.
Meta weighs token caps
Instagram head Adam Mosseri said on Lenny’s Podcast that Meta may need to impose caps on AI token spending per engineer within the next one to two years as internal AI experimentation drives ballooning costs. Mosseri warned that the burn rate of a strong engineer could soon rival their salary or total employment cost, forcing the company to treat AI tokens like finite resources such as payroll or operating expenses. He said the cost of processing AI prompts and responses—measured in tokens—has become a significant operational expense for Meta, and he pointed to the company shutting down an internal AI token spend leaderboard after realizing its AI costs were on track to hit billions of dollars by 2026.
Mosseri argued that Meta should rein in “silly things” that burn tokens without creating value, adding, “It’s not that hard to build a token incinerator, and that doesn’t create a lot of value,”. He drew a direct analogy between token budgets and other operational resources, saying, “Token budgets will be the same.”
Industry shifts on costs
Mosseri said Meta does not currently have token caps for any employee, but he believes their use could become “healthy” as costs grow and as caps are tied to the company’s trust in an engineer’s ability to use the budget in an “ROI-positive” way. He also described how he has to decide how to deploy capacity across teams because Meta has “a limited number of GPUs and CPUs and storage and RAM,” and he said token budgets would be handled alongside those constraints. The same interview framing appeared in coverage that linked Meta’s internal leaderboard shutdown to a broader rethinking of AI experimentation after AI costs put the company on track for billions of dollars in 2026.
Other companies were cited as moving in parallel, including Uber reportedly blowing through its 2026 AI coding budget by April and Microsoft canceling Claude Code licenses while consolidating engineers around its own Copilot CLI tool. In the TechCrunch account, Mosseri said, “I think of it like…any other resource,” and he added, “I have to decide how to deploy capacity to my different teams because I have a limited number of GPUs and CPUs and storage and RAM etc.”
What comes next
Mosseri said any future cap per engineer would be proportional to Meta’s trust in that engineer’s ability to use the budget in an “ROI-positive” way, and he suggested that employees who improve productivity, accelerate development, or generate revenue could receive larger budgets. He also said Meta expects token costs to decline as AI model makers enter a pricing war, but he framed the immediate period as one where tighter controls may be required to prevent runaway spending. Coverage of Meta’s internal cost controls emphasized that the company already reduced token burn by eliminating wasteful projects, including shutting down the token spend leaderboard that ranked employees by usage.
“Meta does not currently impose AI token limits on employees, Mosseri confirmed.”
The TechCrunch account also tied the shift to other “AI reckoning” moments, including Uber’s AI coding budget being exhausted by April and Microsoft canceling Claude Code licenses as token prices rose. Mosseri’s remarks, as quoted in the TechCrunch report, left the company without explicit caps today but pointed toward a future where token budgets become part of how executives allocate resources like payroll and operating expenses, with Mosseri saying, “Token budgets will be the same.”